Does Repeatability Outweigh Accuracy?

In a perfect world, all gaging results are right on. They are both accurate and precise no matter the conditions or by whom or where the gage is used. We wish it were this way, but it’s not. These two gage characteristics are both very important. However, we could ask ourselves which one is more important. Maybe we can make an argument for one over the other.

According to Wikipedia, in the field of metrology, the accuracy of a measurement system is the degree of closeness to which measurements of a quantity approach that quantity’s actual (true) value. The precision of a measurement system—also called reproducibility or repeatability—is the degree to which repeated measurements under unchanged conditions show the same results.

A measurement system can be accurate but not precise, precise but not accurate, neither, or both. Of course, there are degrees of accuracy and precision, and some errors are always going to exist. But sometimes—like getting a haircut and putting on a nice suit for a job interview—repeatability can compensate for a certain degree of inaccuracy.

In the metrology business, we frequently get customer requests for gaging applications that are way out of the ordinary. Either the requirement has to do something against the laws of physics, or it pushes conventional gaging beyond its normal limits. Sometimes, the laws can’t be broken and the gage just can’t be built. But in other cases, by understanding accuracy and precision, there is a chance to move the mountain.

Often the request has to do with air gaging because air is so versatile and “accurate,” especially in the shop environment. However, one of the limiting characteristics of air gaging is its short measuring range, as determined by the physics of the pressure-distance curve. Nevertheless, requests often involve pushing the measuring range well beyond the relatively short range on the linear section of this curve. But air gaging can make the best of a not-so-accurate situation. Here is where repeatability can have the advantage over accuracy.

First, we know that air tooling, being a fixed-plug measuring system, is inherently repeatable. This is because the fixed-plug design virtually limits operator or part influence. So in its normal range, the tooling is very repeatable. And if the diameter is large enough to eliminate the effects of centralization error (or if the jetting can be oriented properly), the tool will be very repeatable, even in the non-linear range of the air system. This is a very good condition because now that we know the system is repeatable, we can apply a number of different correction schemes to make the gage appear—and actually be—more accurate.

If the gage is used primarily as a good/bad measuring system, two possibilities exist. First, the air gaging could be set up as a two-master system with two fixed end points. As long as the part being measured is within these limits, then the part will be good. The gage may not be that accurate with its result, but because the end points are known to be repeatable, it will accurately classify the parts.

Another concept is to set the gage up as a single master and document the actual curve of the measuring system. We know there will be inaccuracies, but it will still repeat. If one side of the curve is accurate but the other side comes up short, the short side could be adjusted to change the tolerance to match the real tolerance as if the gage were accurate. For example, with 0.001-inch input, the air gage normally would display 0.001 inch—as accurate as it should be. But what if we instead input 0.001-inch displacement, and the output comes out to be 0.0008 inch, or 20 percent short? Here we might change the tolerance to 0.0008 inch and use this as the limit of the operator. While it’s not good for data collection and analysis, it would be acceptable for go/no-go part classification.

If data collection and analysis are important, then one equation or correction table could be employed either during or after getting measurement results. In this case, a series of points might be plotted and the correction value added or subtracted to the actual value. The operator would then “read” the corrected value. But this can only happen if the measurement result is repeatable.

In fact, utilizing corrected values is more common than often realized. It is employed in most CMMs, machine tools, length measuring machines and virtually every computer-based measuring system. It’s often just too time-consuming and expensive to build micron accuracy into these systems. But what can be built in, relatively easily, is repeatability. If the system is repeatable, then a correction algorithm can make the gage both accurate and repeatable.

So does this mean that repeatability is the most important characteristic of a measuring system? Well, perhaps not most important, but definitely more flexible.